Granted Patent B2

US 12,726,668 · App. 19/009,721 · Granted Sep 1, 2026

Inventors: Aaron Keith Baughman (Cary, NC); Chandankumar Johakhim Patel (Fairborn, OH); Eduardo Morales (Key Biscayne, FL); Rahul Agarwal (Jersey City, NJ)

Assignee: International Business Machines Corporation

H04N21/25891G06V10/44H04N21/251G06V2201/07

Abstract

Claims (49)

1 . A method comprising:

receiving a plurality of video streams;

performing feature identification to identify objects within scenes of the plurality of video streams;

correlating the identified objects with event outcomes within the video streams;

selecting a first of the video streams with a relatively highest level of correlation to the event outcomes;

diffusing features from a remainder of the video streams into the first video stream to generate a personalized video stream; and

causing the personalized video stream to be transmitted to a user device used by a target user.

2 . The method of claim 1 , wherein the correlating is based on directed acyclic graph structures.

3 . The method of claim 1 , wherein the correlating is based on adjacency matrices across events associated with the event outcomes.

4 . The method of claim 3 , further comprising:

creating a Directed Acyclic Graph (DAG) based on the identified objects and the event outcomes identified within the video streams, wherein the adjacency matrices are based on the DAG.

5 . The method of claim 1 , wherein the diffusing is based on a causal strength associated with the features.

6 . The method of claim 1 , wherein the diffusing is based on prompts to alter the features that are diffused from the remainder of the video streams into the first video stream.

7 . The method of claim 6 , further comprising:

extracting, from a user profile of the target user, information about preferences of the target user; and

causing the information to be analyzed by a trained artificial intelligence (AI) model, wherein an output of the trained AI model includes the prompts.

8 . The method of claim 1 , wherein the performed feature identification is feature pyramidal identification, wherein the feature pyramidal identification is based on a narrowing of features windows to identify the identified objects within frames that the scenes of the plurality of video streams are made up of.

9 . A computer program product comprising:

one or more computer readable storage media; and

program instructions stored on the one or more storage media to perform operations comprising:

receiving a plurality of video streams;

performing feature identification to identify objects within scenes of the plurality of video streams;

correlating the identified objects with event outcomes within the video streams;

selecting a first of the video streams with a relatively highest level of correlation to the event outcomes;

diffusing features from a remainder of the video streams into the first video stream to generate a personalized video stream; and

causing the personalized video stream to be transmitted to a user device used by a target user.

10 . The computer program product of claim 9 , wherein the correlating is based on directed acyclic graph structures.

11 . The computer program product of claim 9 , wherein the correlating is based on adjacency matrices across events associated with the event outcomes.

12 . The computer program product of claim 11 , wherein the operations further comprise:

creating a Directed Acyclic Graph (DAG) based on the identified objects and the event outcomes identified within the video streams, wherein the adjacency matrices are based on the DAG.

13 . The computer program product of claim 9 , wherein the diffusing is based on a causal strength associated with the features.

14 . The computer program product of claim 9 , wherein the diffusing is based on prompts to alter the features that are diffused from the remainder of the video streams into the first video stream.

15 . The computer program product of claim 14 , wherein the operations further comprise:

extracting, from a user profile of the target user, information about preferences of the target user; and

causing the information to be analyzed by a trained artificial intelligence (AI) model, wherein an output of the trained AI model includes the prompts.

16 . The computer program product of claim 9 , wherein the performed feature identification is feature pyramidal identification, wherein the feature pyramidal identification is based on a narrowing of features windows to identify the identified objects within frames that the scenes of the plurality of video streams are made up of.

17 . A computer system comprising:

a processor set;

one or more computer readable storage media; and

program instructions stored on the one or more storage media to cause the processor set to perform operations comprising:

receiving a plurality of video streams;

performing feature identification to identify objects within scenes of the plurality of video streams;

correlating the identified objects with event outcomes within the video streams;

selecting a first of the video streams with a relatively highest level of correlation to the event outcomes;

diffusing features from a remainder of the video streams into the first video stream to generate a personalized video stream; and

causing the personalized video stream to be transmitted to a user device used by a target user.

18 . The computer system of claim 17 , wherein the correlating is based on directed acyclic graph structures.

19 . The computer system of claim 17 , wherein the correlating is based on adjacency matrices across events associated with the event outcomes.

20 . The computer system of claim 17 , wherein the diffusing is based on a causal strength associated with the features.

Assignments (1)

Continuity (1)

References Cited (33)